Re: [net-next] bonding: don't allow the master to become its slave

On Thu, 9 Aug 2012 21:55:39 +0200
Jiri Pirko [off-list ref] wrote:

Thu, Aug 09, 2012 at 09:39:06PM CEST, fbl@redhat.com wrote:

quoted

On Thu, 9 Aug 2012 20:03:23 +0100
Ben Hutchings [off-list ref] wrote:

quoted

On Thu, 2012-08-09 at 15:30 -0300, Flavio Leitner wrote:

quoted

It doesn't make any sense to allow the master to become
its slave. That creates a loop of events causing a crash.

What if there are other intermediate devices, e.g. the slave is a VLAN
sub-device of the bond?  And doesn't team also have this problem?

I think a more general check for such loops might be required.

Maybe patching netdev_set_master() to fail in the loop case is
the way to go.  That would work for bonding, team and bridge.

What you think?

How about other devices who do not use "->master" like vlan, macvlan?

Didn't get you. This is what I had in mind, just to show the idea.

diff --git a/net/core/dev.c b/net/core/dev.c
index f91abf8..a404afb 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c

@@ -4424,6 +4424,24 @@ static int __init dev_proc_init(void)
 #define dev_proc_init() 0
 #endif	/* CONFIG_PROC_FS */
 
+bool netdev_check_loop(struct net_device *master, struct net_device *dev)
+{
+	if (master == dev)
+		return true;
+
+	if (is_vlan_dev(dev))
+		return nedev_check_loop(vlan_dev_real_dev(dev));
+
+	/* is a bridge ?*/
+	if (dev->priv_flags & IFF_EBRIDGE) {
+		list_for_each_entry(p, &br->port_list, list) {
+			if (nedev_check_loop(p->dev))
+				return true;
+		}
+	}
+
+	return false;
+}
 
 /**
  *	netdev_set_master	-	set up master pointer

@@ -4447,6 +4465,9 @@ int netdev_set_master(struct net_device *slave, struct net_device *master)
 		dev_hold(master);
 	}
 
+	if (netdev_check_loop(master, slave))
+		return -EINVAL;
+
 	slave->master = master;
 
 	if (old)

fbl